NCPA - National Center for Policy Analysis

What Caused Pollsters to Misread Voters?

November 8, 2002

The 2002 mid-term election season was not a shining moment for U.S. polling firms. Many of their predictions turned out to be flat wrong. How did this happen?

There are a wide variety of explanations.

  • Twenty years ago, two-thirds or more of Americans were willing to endure the 20 minutes or so that it takes to conduct a telephone interview -- but that proportion has now been cut in half.
  • Many Americans now have caller ID or some other call-blocking mechanism which enables them to screen out calls from pollsters -- and these non-responses can skew results.
  • With the electorate nearly evenly divided between Republicans and Democrats, the inherent margin of error becomes more important.
  • The growth in cellular telephone use hampers efforts to arrive at a representative sample of voters because pollsters use conventional telephone numbers, rather than cell phone numbers, and that distinction precludes reaching many younger voters who are more likely to be cell phone users.

The country's continuing stream of immigrants also makes accurate polling more difficult, since racial and ethnic groups tend to have distinctive voting patterns. It used to be that pollsters were satisfied with representative numbers of whites and blacks. Now in important states such as California and Texas, Hispanic and Asians must be accounted for.

More women in the workforce means they are not at home during the day to answer survey questions. Conversely, retired couples and individuals are more easily reached -- which means older voters may be over-represented in samples.

Source: John Harwood and Shirley Leung, "Political Pollsters Rethink Methods in Wake of Election," Wall Street Journal, November 8, 2002.

For text (WSJ subscribers)


Browse more articles on Government Issues